|
Date |
Topics |
Readings |
Notes |
|
Introduction |
|||
|
Th 9/3 |
Introduction Overview, applications, history |
Bishop 1 |
|
|
Tu 9/8 |
Math Review Probability, stats, linear algebra, optimization |
You do NOT need to read all Probability: Bishop Chapter 2 Bishop Appendix B Tom Minka's nuances of probability (advanced) Linear algebra Bishop Appendix C |
|
|
Th 9/10 |
Foundations from Learning Theory Learning definitions, settings |
Bishop 1 |
|
|
Tu 9/15 |
Decision Trees Construction, pruining, over-fitting |
|
|
|
Supervised Learning: Linear Methods |
|||
|
Th 9/17 |
Regression Least squares and regression |
Bishop 3 |
|
|
Tu 9/22 |
Classification Logistic Regression |
Bishop 4 |
|
|
Th 9/24 |
Generative vs. discriminative Naive Bayes and Logistic Regression |
|
|
|
Tu 9/29 |
Online methods Perceptron |
|
|
|
Supervised Learning: Non-Linear Methods |
|||
|
Th 10/1 |
Support Vector Machines Max-margin classification and optimization |
Bishop 7.1 Bishop Appendix E |
|
|
Tu 10/6 |
Kernel Methods Dual optmization, kernel trick |
Bishop 6.1, 6.2 |
|
|
Th 10/8 |
Instance based learning Nearest-neighbors |
Bishop 2.5 Mitchell 8-8.4 |
|
|
Tu 10/13 |
Neural Networks 1 Neural Network models |
Bishop 5.1,5.2 |
|
|
Th 10/15 |
Neural Networks 2 Learning neural networks |
Bishop 5.3,5.5 |
|
|
Unsupervised Learning |
|||
|
Tu 10/20 |
EM and Clustering 1 Expectation-Maximization and k-means |
Bishop 9 |
|
|
Th 10/22 |
EM and Clustering 2 Gaussian mixture models |
Bishop 9 |
|
|
Tu 10/27 |
Graphical models 1 Bayesian networks and conditional independence |
Bishop 8.1, 8.2 |
|
|
Th 10/29 |
Graphical models 2 MRFs and Exact inference |
Bishop 8.3, 8.4 |
|
|
Complex Output |
|||
|
Tu 11/3 |
Sequential graphical models 1 Max Sum and Max Product |
Bishop 13.1,13.2 |
|
|
Th 11/5 |
Sequential graphical models 2 HMMs and CRFs |
|
|
|
Tu 11/10 |
Dimensionality reduction PCA, probabilistic PCA |
Bishop 12.1,12.2,12.3 |
|
|
Th 11/12 |
Ensemble Methods Boosting and ensembles |
Bishop 14.1,14.2,14.3 |
|
|
Tu 11/17 |
Multi-class Reductions, 1-of-K encoding, structured |
Bishop 4.1.2 Solving Multiclass Learning Problems via Error-Correcting Output Codes (Sections 1, 2.3, 2.4) Reducing Multiclass to Binary (Sections 1, 2, 3) |
|
|
Learning Settings |
|||
|
Th 11/19 |
Learning settings 1 Unsupervised Prediction Aggregation |
|
|
|
Tu 11/24 |
Learning settings 2 Active learning |
|
|
|
Th 11/26 |
Thanksgiving Break No class |
|
|
|
Tu 12/1 |
Learning settings 3 Multi-task learning, transfer learning and domain adaptation |
|
|
|
Th 12/3 |
Learning settings 4 Semi-supervised learning |
|
|
|
Th 12/17 |
Final Exam Time Project presentations 6-9pm |
|
|